NBA MVP Logistical Regression
I’m taking the mlclass from coursera right now and decided to apply a logistical regression classification on nba players to try and determine who will be voted the MVP. The model looked at the common individual stats as well as the team winning percentage. Obviously this model tries to model the voters opinions on who should be the mvp and doesn’t necessarily classify the best player.
I looked at the past 30 years of MVPs and ran the model on each season.
In R, here is a definition of the model:
> mvpmodel
Call: glm(formula = MVP ~ GAMES + PPG + RPG + APG + SPG + BPG + WINPCT,
family = binomial(link = "logit"))
Coefficients:
(Intercept) GAMES PPG RPG APG SPG
BPG WINPCT
-32.919831 0.004281 0.369791 0.365068 0.541685 0.364045
0.505713 19.610476
Degrees of Freedom: 16611 Total (i.e. Null); 16604 Residual
Null Deviance: 562.2
Residual Deviance: 158.2 AIC: 174.2
> After the model was created, I ran it on every player over the past 30 years to find the likelihood that they would be choosen MVP. Then I sorted the list and found the top 10 most dominant MVP-esque seasons over the past 40 years:
1) Kareem Abdul-jabbar 1971 9.805001e-01
2) Shaquille O'neal 1999 9.585589e-01
3) Artis Gilmore 1971 9.331693e-01
4) Kareem Abdul-jabbar 1970 9.279581e-01
5) Larry Bird 1984 8.509291e-01
6) Kareem Abdul-jabbar 1973 8.455859e-01
7) Michael Jordan 1995 8.435103e-01
8) LeBron James 2008 8.339901e-01
9) Magic Johnson 1986 8.267087e-01
10) Michael Jordan 1991 8.180643e-01The only one of those guys to not actually win the MVP was Artis Gilmore. Its a shame because his numbers in 1971 were ridiculous! 24 points 18 rebounds 5 blocks on a team that won 80+% of their games.
Next I ran the model on each year of performances to find out who the model thinks should have been voted for MVP. Then I compared it to who actually was voted MVP. Its kind of interesting because it highlights the years when the voting might have been controversial. Like the steve nash MVPs of recent years.
YEAR ACTUAL_MVP MODEL_PREDICTED_MVP
1970 Kareem Kareem
1971 Kareem Kareem
1972 Dave Cowens Kareem
1973 Kareem Kareem
1974 Bob Mcadoo Julius Erving
1975 Kareem Julius Erving
1976 Kareem Kareem
1977 Bill Walton Bill Walton
1978 Moses Malone Kareem
1979 Kareem Kareem
1980 Julius Erving Julius Erving
1981 Moses Malone Larry Bird
1982 Moses Malone Moses Malone
1983 Larry Bird Larry Bird
1984 Larry Bird Larry Bird
1985 Larry Bird Larry Bird
1986 Magic Johnson Magic Johnson
1987 Michael Jordan Larry Bird
1988 Magic Johnson Magic Johnson
1989 Magic Johnson Michael Jordan
1990 Michael Jordan Michael Jordan
1991 Michael Jordan Michael Jordan
1992 Charles Barkley Charles Barkley
1993 Olajuwon David Robinson
1994 David Robinson David Robinson
1995 Michael Jordan Michael Jordan
1996 Karl Malone Michael Jordan
1997 Michael Jordan Shaq
1998 Karl Malone Karl Malone
1999 Shaq Shaq
2000 Iverson Shaq
2001 Tim Duncan Tim Duncan
2002 Tim Duncan Tim Duncan
2003 Kevin Garnett Kevin Garnett
2004 Steve Nash Dirk Nowitzki
2005 Steve Nash Lebron James
2006 Dirk Nowitzki Dirk Nowitzki
2007 Kobe Bryant Kobe Bryant
2008 Lebron James Lebron James
2009 Lebron James Lebron JamesI only had data up until 2009 for this
Finally I wanted to see if the MVP has changed meaning over the years. Like today the mvp is usually given to the best player on the best team, but in the past maybe it was given to the best player even if they were on a really bad team. If anything, the opposite is true, but really there isn’t much trend to speak of. Here is a chart of the MVP’s team rank over the past 50 years ( 1 is the best ). Also keep in mind the league has added teams over these years.
Chart of MVP team ranks by year
